Will AI actually replace software engineers?
If AI can write code, where does that leave me?
Will AI actually replace software engineers?
You’ve probably heard this a lot lately:
“AI is going to replace software engineers.”
I mean AI is writing code, generating tests, even building small apps end to end, but that statement misses something important.
It treats software engineering as a single problem, but in reality, it’s two very different ones:
The first is building the thing.
The second is deciding what to build in the first place.
Is AI really good at both of them?
Let’s start with where AI is genuinely impressive.
If you give it a clear, well-defined task, it performs extremely well.
“Build an API for this schema / XYZ specifications.”
“Generate tests for this function in ABC way / format.”
“Convert this design into UI components.”etc
At that point, the work becomes very structured, clear, predictable, and unambiguous, and to be honest, AI thrives in such environments.
That’s why a lot of developers feel uneasy - Because a meaningful portion of day-to-day coding does fall into that category.
But… here’s the part that often gets overlooked - That was never the hardest part of the job!!!!
The harder part has always been figuring out what should exist in the first place.
In real-world systems, requirements don’t show up as clean specifications. They instead show up as vague ideas.
Eg: “We should let users delete their account.”
Sounds simple, right?
Is it though?
Let’s start unpacking it:
Do we delete immediately or after a delay?
What happens to data in backups?
What about compliance requirements?
What happens if there’s an active transaction?
How do downstream systems react?
etc, etc, etc, etc
Very quickly, you’re not writing code anymore. You’re making decisions.
And each decision involves tradeoffs like:
Speed vs safety.
Simplicity vs flexibility.
Short-term delivery vs long-term maintenance.
This is where most of the real engineering work lives, and where AI still struggles.
Not because it’s not powerful, but because it lacks context. It:
doesn’t understand your business constraints.
doesn’t sit in conversations with stakeholders (PMs, leadership etc).
doesn’t feel the consequences of a bad decision six months later.
It can generate options, but it can’t reliably choose the right one that provides the best ROI to your team / company - and every business is different! Even if 3 companies are trying to solve the same problem, each will have a different requirement from risk-reward perspective, and there is no 1 size fits all solution.
But something interesting happens at the same time.
As implementation gets easier, decision-making becomes more valuable.
Someone still needs to:
Figure out what problem is actually worth solving.
Clarify ambiguous requirements.
Align teams on tradeoffs.
Anticipate edge cases and future impact.
Decide what “good” looks like.
In fact, these skills become even more important as the cost of building decreases, because when it’s easy to build anything, choosing what to build becomes the bottleneck.
There’s also a second-order effect most people aren’t talking about.
AI increases output, but it can also increase noise:
More code.
More prototypes.
More half-thought-through solutions.
Which means engineers who can create clarity, filter options, and guide decisions become disproportionately valuable - not because they write more code, but because they reduce confusion.
I go deeper into this in the full video, including:
which engineers are most at risk
how AI is actually changing day-to-day work
and what skills will matter even more going forward
If you want the full breakdown, you can watch it here:
So the question isn’t really:
“Will AI replace software engineers?”
A better question is:
“What kind of engineer will still be valuable as AI improves?”
Let me know what you think!